NeuroBridge
AI Enterprise SolutionsGPT & Claude

SignalBridge

Bu layihə bilavasitə şəhər bələdiyyə xidmətləri, telekommunikasiya provayderləri və NOC (Network Operations Center) şöbələri üçün nəzərdə tutulub. Problemin miqyası çox real və ciddidir: son rəqəmsal xidmət statistikalarına əsasən, yerli elektron şikayət sistemlərinə sadəcə 8 ay ərzində 4000-dən çox struktursuz müraciət daxil olur. Hazırda bu qurumlar daxil olan minlərlə mesajı, zəngi və portal müraciətini insan gücü ilə oxuyur və çeşidləyirlər. Nəticədə nə baş verir? Həddindən artıq gecikmələr yaşanır, müştəri məmnuniyyəti (UX indeksi) kəskin düşür və ən təhlükəlisi – bu böyük məlumat axını (səs-küy) içərisində təcili müdaxilə tələb edən kritik infrastruktur qəzaları (məsələn, şəbəkə kəsintiləri) gözərdən qaçır.

Open the demoGitHub

Preliminary score

52.0/ 100Earlyaverage of 2 judges
  • Value for the user15.5 / 25

    A specific problem and a clear outcome from solving it.

  • Prototype and use of AI17.5 / 30

    A working core scenario, and what AI actually contributes to the result.

  • Quality testing8.0 / 20

    Tests, examples of failures, and comparison with the current approach.

  • Feasibility6.5 / 15

    Data requirements, running costs and a clear next step.

  • Originality4.5 / 10

    A meaningful difference from common solutions.

What the judges wrote

  • GPT-6 AstraPreliminary53/100
    Strengths
    Your focus on telecom and NOC complaint triage addresses a concrete operational workflow. Your architecture gives AI a substantive role in interpreting complaints while keeping routing rules explicit. The 0.7 confidence threshold and manual-review fallback for API and schema failures make your intended safeguards concrete.
    To improve
    Show a completed complaint-to-department trace and reproducible results behind your 76/76 test badge, including the injected failure cases. Compare routing accuracy, critical-outage recall, and handling time against manual or keyword triage; deterministic rules do not prevent schema-valid AI misclassifications. Your in-memory ledger needs durable storage and recovery testing before you can support claims of no data loss or 100% continuity. Specify an actual cost per 1,000 signals and a scoped pilot with one telecom team.
  • Claude Opus 5.5Preliminary51/100
    Strengths
    Your project targets a concrete workflow: telecom NOC and municipal complaint triage, grounded in a figure of 4,000+ unstructured requests over 8 months. The architecture is well reasoned. Gemini only extracts sentiment, category, severity and confidence under a JSON schema, while deterministic code makes the routing decision, with a confidence threshold of 0.7. Your circuit breaker, which sends signals to a Manual Review queue on 429 errors, timeouts or schema drift, is a thoughtful reliability choice. The deployed dashboard, with live, cached and offline AI modes, an audit log export and an outage-simulation button, suggests a real working prototype.
    To improve
    Your testing section describes failure scenarios but gives no numbers. Report classification accuracy on a labelled set, the rate of schema-drift failures, and the share of signals falling into Manual Review. Your dashboard shows '76/76 tests' and an 'AI vs Baseline' evaluation tab, so put those results in the submission alongside a comparison with manual sorting time. Replace 'micro-costs per thousand signals' with an actual cost per 1,000 complaints and a concrete pilot step, such as which operator or which ticketing system and over what timeframe. Also align the model you name: the deck and README say Gemini 1.5 Flash, but the screenshot shows gemini-3.0-flash.

The team

  • Efshan Satiyev

    developer

  • Murad AbusatdarLead

    designer

  • Abbas Abusatdarov

    developer